[ https://issues.apache.org/jira/browse/HIVE-6679?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14247192#comment-14247192 ]
Vaibhav Gumashta commented on HIVE-6679: ---------------------------------------- [~thejas] Yes, that was intentional since we already have a max idle time config for http mode. Let me enable the keepalive config by default. Regarding the timeout, do you think couple of minutes should be fine? I think we should keep the default pretty high since we don't want the server to throw {{SocketTimeoutException}} prematurely (note that the timeout doesn't close the socket, it just throws an exception in hitting the timeout on blocking calls). > HiveServer2 should support configurable the server side socket timeout for > all transports types > ----------------------------------------------------------------------------------------------- > > Key: HIVE-6679 > URL: https://issues.apache.org/jira/browse/HIVE-6679 > Project: Hive > Issue Type: Bug > Components: HiveServer2 > Affects Versions: 0.13.0, 0.14.0 > Reporter: Prasad Mujumdar > Assignee: Navis > Fix For: 0.14.1 > > Attachments: HIVE-6679.1.patch.txt, HIVE-6679.2.patch.txt, > HIVE-6679.3.patch > > > HiveServer2 should support configurable the server side socket read timeout > and TCP keep-alive option. Metastore server already support this (and the so > is the old hive server). > We now have multiple client connectivity options like Kerberos, Delegation > Token (Digest-MD5), Plain SASL, Plain SASL with SSL and raw sockets. The > configuration should be applicable to all types (if possible). -- This message was sent by Atlassian JIRA (v6.3.4#6332)